WIP group processors by threads

config/config.dhall

@@ -87,10 +87,11 @@ let Memory =
 let Network =
       { Type = { geometry : PlotGeo_.Type }, default.geometry = PlotGeo_::{=} }
 
+let CoreGroup = { threads : Natural, rows : Natural, padding : Natural }
+
 let Processor =
       { Type =
-          { core_rows : Natural
-          , core_padding : Natural
+          { core_groups : List CoreGroup
           , show_stats : Bool
           , show_plot : Bool
           , table_rows : Natural

src/modules/common.lua

@@ -439,6 +439,14 @@ return function(config)
       )
    end
 
+   M.make_blank_dial = function(x, y, radius, thickness, threshold)
+      return dial.make(
+         geom.make_arc(x, y, radius, DIAL_THETA0, DIAL_THETA1),
+         arc.config(style.line(thickness, CAP_BUTT), patterns.indicator.bg),
+         threshold_indicator(threshold)
+      )
+   end
+
    M.make_dial = function(x, y, radius, thickness, threshold, _format, pre_function)
       return {
          dial = dial.make(

src/modules/graphics.lua

@@ -52,8 +52,14 @@ return function(update_freq, config, common, width, point)
       vid_utilization = 0
    }
 
+   local runtime_status_file = config.dev_power..'/runtime_status'
+
+   local want_nvidia_query = config.show_temp or config.show_clock
+      or config.gpu_util or config.mem_util or config.vid_util
+
    local update_state = function()
-      if i_o.read_file(config.dev_power, nil, '*l') == 'on' then
+      local is_active = i_o.read_file(runtime_status_file, nil, '*l') == 'active'
+      if is_active and want_nvidia_query then
          local nvidia_settings_glob = i_o.execute_cmd(NV_QUERY)
          if nvidia_settings_glob == nil then
             mod_state.error = 'Error'
@@ -68,6 +74,8 @@ return function(update_freq, config, common, width, point)
                = __string_match(nvidia_settings_glob, NV_REGEX)
             mod_state.error = false
          end
+      elseif is_active then
+         mod_state.error = false
       else
          mod_state.error = 'Off'
       end

src/modules/processor.lua

@@ -1,3 +1,4 @@
+local dial = require 'dial'
 local compound_dial = require 'compound_dial'
 local text_table = require 'text_table'
 local i_o = require 'i_o'
@@ -21,7 +22,9 @@ return function(update_freq, main_state, config, common, width, point)
    -----------------------------------------------------------------------------
    -- processor state
 
-   local mod_state = cpu.read_cpu_loads(cpu.init_cpu_loads())
+   local topology = cpu.get_core_topology()
+   local mod_state = cpu.read_cpu_loads(cpu.init_cpu_loads(topology))
+   local ncpus = cpu.get_cpu_number(topology)
 
    local update_state = function()
       mod_state = cpu.read_cpu_loads(mod_state)
@@ -30,35 +33,39 @@ return function(update_freq, main_state, config, common, width, point)
    -----------------------------------------------------------------------------
    -- cores (loads and temps)
 
-   local ncpus = cpu.get_cpu_number()
-   local ncores = cpu.get_core_number()
-   local nthreads = ncpus / ncores
+   -- TODO add this back
+   -- local is_evenly_distributed = function(ncores, rows)
+   --    if rows == 0 then
+   --       return false
+   --    elseif math.fmod(ncores, rows) == 0 then
+   --       return true
+   --    else
+   --       i_o.warnf('could not evenly distribute %i cores over %i rows', ncores, rows)
+   --       return false
+   --    end
+   -- end
 
-   local show_cores = false
-
-   if config.core_rows > 0 then
-      if math.fmod(ncores, config.core_rows) == 0 then
-         show_cores = true
-      else
-         i_o.warnf(
-            'could not evenly distribute %i cores over %i rows; disabling',
-            ncores,
-            config.core_rows
-         )
-      end
-   end
-
-   local create_core = function(core_cols, y, c)
+   local create_core = function(core_cols, y, nthreads, padding, c)
       local dial_x = point.x +
          (core_cols == 1
           and (width / 2)
-          or (config.core_padding + dial_outer_radius +
-              (width - 2 * (dial_outer_radius + config.core_padding))
+          or (padding + dial_outer_radius +
+              (width - 2 * (dial_outer_radius + padding))
               * math.fmod(c - 1, core_cols) / (core_cols - 1)))
       local dial_y = y + dial_outer_radius +
          (2 * dial_outer_radius + dial_y_spacing)
          * math.floor((c - 1) / core_cols)
-      return {
+      local loads
+      if nthreads == 1 then
+         local single_thickness = dial_outer_radius - dial_inner_radius
+         loads = common.make_blank_dial(
+            dial_x,
+            dial_y,
+            dial_outer_radius - single_thickness / 2,
+            single_thickness,
+            80
+         )
+      else
          loads = common.make_compound_dial(
             dial_x,
             dial_y,
@@ -67,7 +74,10 @@ return function(update_freq, main_state, config, common, width, point)
             dial_thickness,
             80,
             nthreads
-         ),
+         )
+      end
+      return {
+         loads = loads,
          coretemp = common.make_text_circle(
             dial_x,
             dial_y,
@@ -79,44 +89,65 @@ return function(update_freq, main_state, config, common, width, point)
       }
    end
 
-   local mk_cores = function(y)
-      local core_cols = ncores / config.core_rows
-      local cores = pure.map_n(pure.partial(create_core, core_cols, y), ncores)
-      local coretemp_paths = cpu.get_coretemp_paths()
-      if #coretemp_paths ~= ncores then
-         i_o.warnf('could not find all coretemp paths')
-      end
-      local update_coretemps = function()
-         for conky_core_idx, path in pairs(coretemp_paths) do
-            local temp = __math_floor(0.001 * i_o.read_file(path, nil, '*n'))
-            common.text_circle_set(cores[conky_core_idx].coretemp, temp)
-         end
-      end
-      local update = function()
-         for _, load_data in pairs(mod_state) do
-            compound_dial.set(
-               cores[load_data.conky_core_idx].loads,
-               load_data.conky_thread_id,
-               load_data.percent_active * 100
+   local mk_core_group = function(group_config, y)
+      local nthreads = group_config.threads
+      local core_topology = topology[nthreads]
+      local ncores = #core_topology
+      local core_cols = ncores / group_config.rows
+      local _create_core = pure.partial(
+         create_core, core_cols, y, nthreads, group_config.padding
+      )
+      local cores = pure.map_n(_create_core, ncores)
+      local group_loads = mod_state[nthreads]
+      local update_loads
+      local draw_static_loads
+      local draw_dynamic_loads
+      if nthreads == 1 then
+         update_loads = function(c)
+            dial.set(
+               cores[c].loads,
+               group_loads[c][1].percent_active * 100
             )
          end
-         update_coretemps()
+         draw_static_loads = dial.draw_static
+         draw_dynamic_loads = dial.draw_dynamic
+      else
+         update_loads = function(c)
+            for t = 1, nthreads do
+               compound_dial.set(
+                  cores[c].loads,
+                  t,
+                  group_loads[c][t].percent_active * 100
+               )
+            end
+         end
+         draw_static_loads = compound_dial.draw_static
+         draw_dynamic_loads = compound_dial.draw_dynamic
+      end
+      local update = function()
+         for c = 1, ncores do
+            local temp = __math_floor(
+               0.001 * i_o.read_file(core_topology[c].coretemp_path, nil, '*n')
+            )
+            common.text_circle_set(cores[c].coretemp, temp)
+            update_loads(c)
+         end
       end
       local static = function(cr)
-         for i = 1, #cores do
+         for i = 1, ncores do
             common.text_circle_draw_static(cores[i].coretemp, cr)
-            compound_dial.draw_static(cores[i].loads, cr)
+            draw_static_loads(cores[i].loads, cr)
          end
       end
       local dynamic = function(cr)
-         for i = 1, #cores do
+         for i = 1, ncores do
             common.text_circle_draw_dynamic(cores[i].coretemp, cr)
-            compound_dial.draw_dynamic(cores[i].loads, cr)
+            draw_dynamic_loads(cores[i].loads, cr)
          end
       end
       return common.mk_acc(
          width,
-         (dial_outer_radius * 2 + dial_y_spacing) * config.core_rows
+         (dial_outer_radius * 2 + dial_y_spacing) * group_config.rows
          - dial_y_spacing,
          update,
          static,
@@ -172,8 +203,12 @@ return function(update_freq, main_state, config, common, width, point)
       )
       local update = function()
          local s = 0
-         for i = 1, #mod_state do
-            s = s + mod_state[i].percent_active
+         for g = 1, #mod_state do
+            for c = 1, #mod_state[g] do
+               for t = 1, #mod_state[g][c] do
+                  s = s + mod_state[g][c][t].percent_active
+               end
+            end
          end
          common.tagged_percent_timeseries_set(total_load, s / ncpus * 100)
       end
@@ -225,14 +260,18 @@ return function(update_freq, main_state, config, common, width, point)
    -----------------------------------------------------------------------------
    -- main functions
 
+   local core_group_section = function (g)
+      return {pure.partial(mk_core_group, g), true, text_spacing}
+   end
+
    return {
       header = 'PROCESSOR',
       point = point,
       width = width,
       set_state = update_state,
       top = {
-         {mk_cores, show_cores, text_spacing},
-         {mk_hwp_freq, config.show_stats, sep_spacing},
+         table.unpack(pure.map(core_group_section, config.core_groups)),
+         -- {mk_hwp_freq, config.show_stats, sep_spacing},
       },
       common.mk_section(
          sep_spacing,

src/pure.lua

@@ -189,6 +189,32 @@ M.flatten = function(xs)
    return r
 end
 
+M.group_with = function(keyfun, valfun, seq)
+   local f = function(acc, next)
+      local k = keyfun(next)
+      local v = valfun(next)
+      if acc[k] == nil then
+         acc[k] = {v}
+      else
+         acc[k][#acc[k] + 1] = v
+      end
+      return acc
+   end
+   return M.reduce(f, {}, seq)
+end
+
+M.group_by = function(k, seq)
+   local f = function(acc, next)
+      if acc[k] == nil then
+         acc[k] = {next}
+      else
+         acc[k][#acc[k]] = next
+      end
+      return acc
+   end
+   return M.reduce(f, {}, seq)
+end
+
 M.concat = function(...)
    return M.flatten({...})
 end

src/sys.lua

@@ -178,8 +178,15 @@ M.get_core_number = function()
    return __tonumber(i_o.read_file('/proc/cpuinfo', 'cpu cores%s+:%s(%d+)'))
 end
 
-M.get_cpu_number = function()
-   return __tonumber(i_o.execute_cmd('nproc', nil, '*n'))
+M.get_cpu_number = function(topology)
+   local n = 0
+   for g = 1, #topology do
+      for c = 1, #topology[g] do
+         n = n + #topology[g][c].cpus
+      end
+   end
+   return n
+   -- return __tonumber(i_o.execute_cmd('nproc', nil, '*n'))
 end
 
 local get_coretemp_dir = function()
@@ -188,27 +195,123 @@ local get_coretemp_dir = function()
    return pure.fmap_maybe(dirname, s)
 end
 
--- map cores to integer values starting at 1; this is necessary since some cpus
--- don't report their core id's as a sequence of integers starting at 0
-local get_core_id_indexer = function()
+-- return a table with keys corresponding to physcial core id and values to
+-- the number of threads of each core (usually 1 or 2)
+M.get_core_threads = function()
+   local cmd = 'lscpu -y -p=core | grep -v \'^#\' | sort -k1,1n | uniq -c'
+   local flip = function(c) return {__tonumber(c[2]), __tonumber(c[1])} end
    local make_indexer = pure.compose(
       pure.array_to_map,
-      pure.partial(pure.imap, function(i, c) return {__tonumber(c), i} end),
-      pure.partial(gmatch_to_table1, '(%d+)')
-   )
-   return pure.fmap_maybe(
-      make_indexer,
-      i_o.execute_cmd('lscpu -p=CORE | tail -n+5 | sort | uniq')
+      pure.partial(pure.map, flip),
+      pure.partial(gmatch_to_tableN, '(%d+) (%d+)')
    )
+   return pure.fmap_maybe(make_indexer, i_o.execute_cmd(cmd))
 end
 
+
+-- map cores to integer values starting at 1; this is necessary since some cpus
+-- don't report their core id's as a sequence of integers starting at 0
+-- local get_core_id_indexer = function()
+--    local make_indexer = pure.compose(
+--       pure.array_to_map,
+--       pure.partial(pure.imap, function(i, c) return {__tonumber(c), i} end),
+--       pure.partial(gmatch_to_table1, '(%d+)')
+--    )
+--    return pure.fmap_maybe(
+--       make_indexer,
+--       i_o.execute_cmd('lscpu -p=CORE | tail -n+5 | sort -k1,1n')
+--    )
+-- end
+
+-- conky_core_idx: the ID of the dial to be drawn for this core
+-- conky_thread_idx: the ID of the individual indicator within one dial
+--   corresponding to one thread in a core (starting at 1 for each core)
+
+local get_coretemp_mapper = function()
+   local d = get_coretemp_dir()
+   i_o.assert_exe_exists('grep')
+   local get_labels = pure.compose(
+      i_o.execute_cmd,
+      pure.partial(__string_format, 'grep Core %s/temp*_label', true)
+   )
+   local to_tuple = function(m)
+      return {__tonumber(m[2]), __string_format('%s/%s_input', d, m[1])}
+   end
+   local to_map = pure.compose(
+      pure.array_to_map,
+      pure.partial(pure.map, to_tuple),
+      pure.partial(gmatch_to_tableN, '/([^/\n]+)_label:Core (%d+)\n')
+   )
+   return pure.maybe({}, to_map, pure.fmap_maybe(get_labels, d))
+end
+
+M.get_core_topology = function()
+   local coretemp_paths = get_coretemp_mapper()
+   local assign_cpu = function(i, x)
+      return {
+         lgl_cpu_id = i,
+         phy_core_id = __tonumber(x[1]),
+         phy_cpu_id = __tonumber(x[2])
+      }
+   end
+   local assign_core = function(acc, next)
+      local g = acc.grouped
+      local max_lgl_core_id = #g
+      local new_phy_core_id = next.phy_core_id
+      local new_cpu = {phy_cpu_id = next.phy_cpu_id, lgl_cpu_id = next.lgl_cpu_id}
+      if acc.prev_phy_core_id == new_phy_core_id then
+         local max_thread = #acc.grouped[max_lgl_core_id].cpus
+         acc.grouped[max_lgl_core_id].cpus[max_thread + 1] = new_cpu
+      else
+         local new_lgl_core_id = max_lgl_core_id + 1
+         acc.grouped[new_lgl_core_id] = {
+            phy_core_id = new_phy_core_id,
+            lgl_core_id = new_lgl_core_id,
+            coretemp_path = coretemp_paths[new_phy_core_id],
+            cpus = {new_cpu}
+         }
+         acc.prev_phy_core_id = new_phy_core_id
+      end
+      return acc
+   end
+   local get_threads = function(x)
+      return #x.cpus
+   end
+   local f = pure.compose(
+      pure.partial(pure.group_with, get_threads, pure.id),
+      pure.partial(pure.get, 'grouped'),
+      pure.partial(pure.reduce, assign_core, {prev_phy_core_id = -1, grouped = {}}),
+      pure.partial(pure.imap, assign_cpu),
+      pure.partial(gmatch_to_tableN, '(%d+),(%d+)')
+   )
+   local out =
+      i_o.execute_cmd('lscpu -y -p=core,cpu | grep -v \'^#\' | sort -k1,1n')
+   return pure.fmap_maybe(f, out)
+end
+
+-- for t, k in pairs(get_core_topology()) do
+--    print(t)
+--    for x, y in pairs(k) do
+--       print(x, y.phy_core_id, y.coretemp_path, #y.cpus)
+--       -- for _, z in pairs(y.cpus) do
+--       --    print(x,z.cpu,z.conky_cpu)
+--       -- end
+--    end
+-- end
+
 local get_core_mappings = function()
-   local ncores = M.get_core_number()
+   local core_threads = M.get_core_threads()
+   local assign_cpus = function(x)
+      return {
+         cpu_id = __tonumber(x[1]),
+         core_id = __tonumber(x[2])
+      }
+   end
    local map_ids = function(indexer)
       local f = function(acc, next)
          local cpu_id = __tonumber(next[1]) + 1
-         local core_id = next[2]
-         local conky_core_idx = indexer[__tonumber(core_id)]
+         local core_id = __tonumber(next[2])
+         local conky_core_idx = indexer[core_id]
          acc.mappings[cpu_id] = {
             conky_core_idx = conky_core_idx,
             conky_thread_id = acc.thread_ids[conky_core_idx],
@@ -219,12 +322,12 @@ local get_core_mappings = function()
       local cpu_to_core_map = pure.maybe(
          {},
          pure.partial(gmatch_to_tableN, '(%d+),(%d+)'),
-         i_o.execute_cmd('lscpu -p=cpu,CORE | tail -n+5')
+         i_o.execute_cmd('lscpu -y -p=cpu,core | grep -v \'^#\' | sort -k1,1n')
       )
-      local init = {mappings = {}, thread_ids = pure.rep(ncores, 1)}
+      local init = {mappings = {}, _conky_core_index = 0, _thread_ids = {}}
       return pure.reduce(f, init, cpu_to_core_map).mappings
    end
-   return pure.fmap_maybe(map_ids, get_core_id_indexer())
+   -- return pure.fmap_maybe(map_ids, )
 end
 
 M.get_coretemp_paths = function()
@@ -306,35 +409,49 @@ M.read_hwp = function(hwp_paths)
    return mixed and 'Mixed' or (HWP_MAP[hwp_pref] or 'Unknown')
 end
 
-M.init_cpu_loads = function()
-   local m = get_core_mappings()
+M.init_cpu_loads = function(topo)
+   -- -- local m = get_core_mappings()
+   -- local topo = get_core_topology()
    local cpu_loads = {}
-   for cpu_id, core in pairs(m) do
-      cpu_loads[cpu_id] = {
-         active_prev = 0,
-         total_prev = 0,
-         percent_active = 0,
-         conky_core_idx = core.conky_core_idx,
-         conky_thread_id = core.conky_thread_id,
-      }
+   for core_group_id, core_group in pairs(topo) do
+      cpu_loads[core_group_id] = {}
+      for lgl_core_id, core in pairs(core_group) do
+         cpu_loads[core_group_id][lgl_core_id] = {}
+         for thread_id = 1, #core.cpus do
+            cpu_loads[core_group_id][lgl_core_id][thread_id] = {
+               active_prev = 0,
+               total_prev = 0,
+               percent_active = 0,
+               -- core_id = lgl_core_id,
+               -- thread_id = thread_id,
+            }
+         end
+      end
    end
    return cpu_loads
 end
 
 M.read_cpu_loads = function(cpu_loads)
-   local ncpus = #cpu_loads
    local iter = io.lines('/proc/stat')
    iter() -- ignore first line
-   for i = 1, ncpus do
-      local ln = iter()
-      local user, system, idle = __string_match(ln, '%d+ (%d+) %d+ (%d+) (%d+)', 4)
-      local active = user + system
-      local total = active + idle
-      local c = cpu_loads[i]
-      if total > c.total_prev then -- guard against 1/0 errors
-         c.percent_active = (active - c.active_prev) / (total - c.total_prev)
-         c.active_prev = active
-         c.total_prev = total
+   for group_id = 1, #cpu_loads do
+      local group = cpu_loads[group_id]
+      for lgl_core_id = 1, #group do
+         local core = group[lgl_core_id]
+         for thread_id = 1, #core do
+            local ln = iter()
+            local user, system, idle =
+               __string_match(ln, '%d+ (%d+) %d+ (%d+) (%d+)', 4)
+            local active = user + system
+            local total = active + idle
+            local thread = core[thread_id]
+            if total > thread.total_prev then -- guard against 1/0 errors
+               thread.percent_active =
+                  (active - thread.active_prev) / (total - thread.total_prev)
+               thread.active_prev = active
+               thread.total_prev = total
+            end
+         end
       end
    end
    return cpu_loads